The Journal of Neuroscience, January 1993, 13(l): 208-229

Rapid Evolution of the Visual System: A Cellular Assay of the Retina and Dorsal Lateral Geniculate Nucleus of the Spanish Wildcat and the Domestic Cat

Robert W. Williams,’ Carmen Cavada,2 and Fernando Reinoso-Suhrez* ‘Department of Anatomy and Neurobiology, College of Medicine, University of Tennessee, Memphis, Tennessee 38163 and *Departamento de Morfologia, Facultad de Medicina, Universidad Aut6noma de Madrid, 28029 Madrid, Spain

The large Spanish wildcat, Fe/is silvestris tartessia, has re- and important topic, it has been difficult to study the process tained features of the Pleistocene ancestor of the modern of brain evolution in any detail. Our approach has been to domestic cat, F. catus. To gauge the direction and magnitude identify a pair of closely related living species,one from a highly of short-term evolutionary change in this lineage, we have conservative branch that has retained near identity with the compared the retina, the optic nerve, and the dorsal lateral ancestral species,and the other from a derived branch that has geniculate nucleus (LGN) of Spanish wildcats and their do- undergone rapid evolutionary change. The recent recognition mestic relatives. Retinas of the two species have the same that evolution and speciationcan occur in short bursts separated area. However, densities of cone photoreceptors are higher by long interludes of stasisprovides a sound theoretical basis in wildcat-over 100% higher in the area centralis-where- for a search for such pairs (Schindewolf, 1950; Eldredge and as rod densities are as high, or higher, in the domestic lin- Gould, 1972; Stanley, 1979; Gould and Eldredge, 1986). eage. Densities of retinal ganglion cells are typically 20- A candidate pair among mammals is the wildcat, Fe/is sil- 100% higher across the wildcat retina, and the total ganglion vestris,and the domestic cat, F. catus. Historical, biochemical, cell population is nearly 70% higher than in the domestic and genetic data point to a derivation of the domestic cat from cat. These differences are confined to the populations of the North African wildcat, F. silvestris lybica (Darwin, 1890; beta and gamma retinal ganglion cells. In contrast, the pop- Clutton-Brock, 1981; Benveniste, 1985; Collier and O’Brien, ulation of alpha cells is almost precisely the same in both 1985; Wayne et al., 1989). This small North African wildcat species. The wildcat LGN is much larger than that of the descendedfrom a larger ancestorwhose Pleistocene fossil record domestic cat and contains-50% more neurons. However, has been describedcarefully by Km-ten (1965a,b). These Pleis- cell size does not differ appreciably in either the retina or tocene wildcats share a constellation of features-short pre- LGN of these species. The differences in total numbers of molar-canine diastema, large canine and carnassialteeth, and ganglion cells and LGN neurons correspond neatly to the a low crowned fourth lower premolar-with the particular sub- overall decline in brain size in the domestic lineage and to specieswe have studied, F. silvestristartessia (Km-ten, 1965a,b). allometric predictions based on average species differences In fact, the Spanishwildcat has more in common with wildcats in body size. We suggest that an increase in the severity of living toward the end of the Wtirm glaciation, 15,000-20,000 naturally occurring cell death is the most plausible mecha- years before present (BP) than with other extant F. silvestris nism that can account for the rapid evolutionary reduction subspecies(Table 1). For this reason,KurtCn (1965b)concluded in cell populations in this feline lineage. that the Spanish subspeciesis an isolated relict of an ancient [Key words: brain size, allometry, rods and cones, retinal Mediterranean wildcat population that has survived with min- ganglion cell, optic nerve, lateral geniculate nucleus, cell imal changefor over 20,000 years in a warm Iberian refuge. death] We decided that a comparison of this particular wildcat sub- specieswith the domestic cat could provide detailed insight into The tempo of mammalian brain evolution has beenrapid. Two- the range, rate, and direction of evolutionary changein brain fold changesin brain weight have occurred over periods of l- structure. For example, we thought we would be able to deter- 5 million years in several lineages,including that of humans mine whether the rapid decline in body size in the cat lineage (Edinger, 1948, 1966; Jerison, 1973, 1979; Radinsky, 1973, has been associatedwith a matched decline in brain massand, 1975, 198 1; Eisenberg, 1981). Although this is an interesting if so, whether this changehas been brought about by a reduction in neuron number or an increase in cell packing density. We Received Apr. 14, 1992; revised July 13, 1992; accepted July 15, 1992. also thought it would be possibleto determine whether numer- This research was supported in part by Grant DGICYT PB88-0170 and a grant ical relations among interconnected parts of the brain have been from the NIH. We are especially indebted to Mariano Sanz Pech and Antonio conserved in this rapidly evolving mammalian lineage, and if Ptrez Alonso-Geta ofthe Institute for the Conservation ofNature, Madrid, Spain, so, then how precisely. Finally, we thought it might be possible for wildcat material. We thank Drs. Leo Chalupa, Andrea Elberger, and Douglas Frost for the generous use of material from their collections of domestic cat tissue, to compare ontogenetic differencesin thesespecies and thereby and Drs. Leo Chalupa, Robert Foehring, Dan Goldowitz, Todd Preuss, Anton exposedevelopmental and cellular mechanismsthat ultimately Reiner, and Michael Rowe for critical comments on drafts of the manuscript. Correspondence should be addressed to Dr. Robert Williams at the above underlie phyletic change. address. Taking advantage of the wealth of data on the structure of Copyright 0 1993 Society for Neuroscience 0270-6474/93/130208-21$05.00/O the primary visual systemof domestic cats, we have begunwith The Journal of Neuroscience, January 1993, 13(i) 209 a quantitative analysis of populations of cells in the retina and Femurs of European wildcats are typically less than 13 cm long (Rijhrs, the dorsal lateral geniculate nucleus (LGN) of the Spanish wild- 1955). HRP injection procedure. We made injections of horseradish perox- cat. In this analysis we make two explicit assumptions: first, idase (HRP) into the left dorsal lateral geniculate nucleus (LGN), left that th; adult wildcats, which we have had the very rare op- superior colliculus, and left pretectal region of the male wildcat (SI).

portunitv to studv,_ I are representative of the Spanish subspecies; Similar unilateral (n = 1) and bilateral (n = 3) series of injections were and second, that the Spanish wildcat is itselfrepresentative of, made into normal domestic cats. Some of the domestic cat retinas have or at least similar to, the ancestral population of Pleistocene been described and illustrated in previous studies (Chalupa et al., 1984; Kirby and Chalupa, 1986). Animals initially received a single intra- wildcats from which domestic cats ultimately trace their descent. muscular dose of ketamine f0.4~ cc)I and atrooine (0.2 cc) and were The validity of these key assumptions is supported by data subseauentlv anesthetized with an intravenous miection of sodium pen- summarized in the next section (Table 1). From this vantage tobarbital (30 mg/kg). They were placed in a stereotaxic device-and point, we attempt to assess structural changes in the primary craniotomies were made over the approximate locations of the left superior colliculus, pretectum, and dorsal lateral geniculate nucleus. The visual system associated with approximately 25,000 years of boundaries of retinorecipient nuclei were mapped by recording multi- natural selection at the close of the Pleistocene, followed by cellular visually evoked activity with a tungsten microelectrode..A series 3000-4000 years of domestication. of 150-200 nl iniections of 25% HRP diluted in oure dimethvl sulfoxide (DMSO) were made along each penetration within the LGN (23 pen- etrations), pretectum (6 penetrations), and superior colliculus (10 pen- etrations) as described in Chalupa et al. (1984). The total volume of Materials and Methods HRP delivered in a single penetration ranged from 200 to 600 nl. We received two adult wildcats (Felis silvestris tartessiu Miller), a male Perfusion and dissection of the animals. All animals received an over- (SI) and a female (S2), from the Institute for the Conservation ofNature dose of pentobarbital(200 mg) and were subsequently perfused through (ICONA), Spain. Both animals had been rescued from game traps in the heart with phosphate-buffered saline (0.1 M, pH 7.3), followed by a the Sierra Morena of south central Spain. The two thin but healthy mixture of 1.25% paraformaldehyde and 2.5% glutaraldehyde in phos- animals had been held for several months in a large outdoor pen. Breed- phate buffer. We removed the cranial vault in one large intact piece, ing was thought to have failed. Both animals had the coat color and and the brain-including olfactory bulb, pituitary, and the entire me- pattern typical of wildcats and many domestic tabby cats (Haltenorth, dulla (Fig. I)-the eyes, and segments of most cranial nerves were re- 1953; Rodriguez de la Fuente, 1979). Each had four prominent dark moved. We took retinas from the eyes and flattened them by making a streaks running down the neck, a single broad dark band running along set of radial cuts. The areas of flat-mounted retinas were measured before the dorsal midline to the base of the tail, six to eight less distinct dark dehydration. We cleaned crania and dentaries in a 25% solution of arched bands running from back to abdomen, and three to four dark bleach. Pelts of both wildcats were returned to ICONA. rings around limbs and tails. Figure 26 in Haltenorth (1953) shows a Following the transcardial perfusion of wildcat S2, an adult female specimen of the F. silvestris tartessia type at the Zoologische Staats- that we had palpated on two previous occasions to be certain that she sammlung (Munich, Germany) with a coat pattern almost indistinguish- was not pregnant, we explored the abdominal cavity and found a single able from that of our animals. The ages of the two wildcats were un- fetus. The fetus was immediately removed and fixed briefly by transcar- known, but both appeared to be young adults. A single fetus was found dial perfusion followed by prolonged immersion in fixative. Preserva- in the female after her perfusion. tion of optic nerves proved to be surprisingly good, and given the value F. silvestris is not listed as an endangered or threatened species under of this rare fetal wildcat material, we undertook a detailed analysis of the U.S. Endangered Species Act or the Convention for International both nerves using procedures identical to those described in a previous Trade in Endangered Species of Wild Fauna and Flora (CITES, 199 1). study on the optic nerves of domestic cat embryos (Williams et al., However, all Felidae except the domestic cat, F. cutus, are listed in 1986). Appendix 2 of CITES and are subject to import and export regulations. Processing of retinas. A modification of the Hanker-Yates procedure The wildcat is protected by law in Spain. We complied with all con- (Perry and Linden, 1982; Chalupa et al., 1984) was used to demonstrate ditions of the ICONA permit and all relevant CITES regulations. the presence of HRP within retinal ganglion cells in wildcat SI and Conjirmation of the wildcat subspecies and the morphological simi- normal domestic cat retinas. We stained retinas of the female wildcat laritv between Soanish wildcats and Pleistocene wildcats. The size and with toluidine blue using the method described by Wong and Hughes morphology of the teeth and jaws are among the most reliable criteria (1987b). Additional domestic cat retinas were mounted between cov- by which to identify species and subspecies of vertebrates (Kurten, 1954; erslips in glycerin and used for the analysis of the photoreceptor mosaic Carroll, 1988). Without resorting to statistical analysis, the dental mea- using methods summarized in Williams (199 1). surements taken from our specimens conform closely to those of three Bruin histology. Brains were hemisected in the sagittal plane. The left F. silvestris tartessia specimens previously studied by Kurten (1965a; hemispheres were cut frozen at 50 pm in the coronal plane. These frozen Table 1). What is even more remarkable and significant is that our sections were collected and distributed in several series. The first and wildcats also do not differ appreciably from a sample of Pleistocene third series were stained with cresyl violet; the second series was in wildcats dated = 15,000 years before present (BP, late Main Wtirm). A some cases reacted for HRP histochemistry using diaminobenzidine as relatively objective index of the degree of similarity or dissimilarity is the chromogen. Right hemispheres were embedded in celloidin and cut orovided bv the t value calculated for all 13 parameters (Table 1, far in either the coronal or the horizontal plane at 50 pm. For comparison, right). This-more rigorous comparison demonstrates that our Spanish frozen and celloidin coronal sections from 11 domestic cats were ex- wildcats are remarkablv similar to Late Main Wihm wildcats (t = 0.64). amined. but differ greatly from-modem European wildcats (t = 7.1 l),‘from do- Photoreceptor analysis. We measured the density of rods and cones mestic cats, including a huge 9 kg domestic tabby cat (t = 4.76), and along the horizontal meridian in the left retina of SI and two domestic from a sizable sample of Neolithic (7500 BP) wildcats (t = 2.41). The cats. A 63 x 1.25 NA semiapochromatic objective with a very long overall body size of our wildcats also overlaps measurements on F. working distance (500 pm) was used to focus on the photoreceptor silvestris tartessia provided by Haltenorth (1953, his Tables 2 and 4). mosaic from either the vitreal or scleral side of the retina using differ- In sum, the animals we have studied have been correctly classified as ential interference contract (DIC) optics, a video overlay system, and F. silvestris tartessia. They are certainly not feral domestic cats, wild semiautomatic image analysis software written for use with an Apple hybrids, or a different subspecies of F. silvestris. These data also provide Macintosh computer (Fig. 2). A 50 W high-pressure mercury light source a strong case that F. silvestris tartessia represents an isolated relict of a was used with or without narrow-band interference filters to obtain late Pleistocene wildcat population. optimal contrast and resolution. Analog and digital video enhancement Sources ofdomestic cat material. Domestic cat tissue was obtained were used to improve the contrast ofthe photoreceptor mosaic as viewed from our colonies in Madrid and the University ofTennessee, and from on an RGB monitor. Rods and cones were counted in fields of 144 Nrn* Drs. L. M. Chalupa, A. Elberger, and D. Frost. An extremely large and 1000 pm*, respectively, using a video overlay setup (Wikler et al., domestic cat was provided to us by Dr. A. Tuberville to explore the 1990; Williams, 199 1). Locations of sites in the retina were measured upper limits in the size of the brain and dentary that might be expected with reference to the center of the area centralis using linear encoders among domestic cats. This huge young domestic cat weighed 9 kg, was attached to the stage (Heidenhain Inc., LS 403 encoders; 0.2 pm reso- 61 cm from crown to rump, and had femurs that were 16.5 cm long. lution; accuracy, * 10 pm over 50 mm). 210 Williams et al. l Rapid Evolution of the Visual System

Table 1. Size of teeth used to assess status of wildcats

Ci P,

Felis type Length Width Length Width Height Felis cutus (Spanish sample) range 4.5-5.6 3.24.1 4.7-5.3 2.3-2.4 3.5-4.0 mean (n = 3) 5.00 3.57 5.08 2.38 3.75 F. silvestris tartessia (southern Spain) SI male (present study) 6.10 4.75 6.45 2.95 4.40 S2 female (present study) 5.20 4.10 6.15 2.90 4.35 mean (n = 2) 5.65 4.43 6.30 2.93 4.38 F. silvestris tartessia (southern Spain) range 5.3-5.6 4.2-4.4 6.1-6.8 3.0-3.1 3.8-4.5 modern< mean (n = 2-3) 5.50 4.33 6.37 3.07 4.33 F. silvestris silvestris (European sample) range 3.6-5.6 3.1-4.3 5.3-6.2 2.5-3.1 3.8-4.5 modern< mean (n = 16-23) 4.52 3.51 5.73 2.83 4.17 F. silvestris (Palestinian sample) mean (n = 10-21) 4.68 3.65 5.44 3.00 4.40 postglacial (- 7600 BP)” F. silvestris (Palestinian sample) mean (n = 5-10) 5.98 4.23 6.41 3.12 4.54 late Main Wiirm (- 15,000 BP)< d F. silvestris (Palestinian sample) mean (n = 6-13) 5.70 4.40 6.64 3.33 4.72 early Main Wiirm (-25,000 BP)d All data are in millimeters. Measurements were taken from teeth ofthe lower jaw. ( Data source; Kurt& (1965b). Ci, canine; P, and P,, the two premolars; M,, the mandibular molar tooth, Protoc., d Data source; Kurt.% (1965a). the protoconid cusp; Parac., the paraconid cusp. nEleven to 13 parameters were compared to generate the t scores. h For testsof significance, degrees of freedom was typically set at a conservative value (n of teeth - I).

Retinal ganglion cell analysis. The analysis of the distribution and center cells have dendritic arbors in the inner and outer parts of the types of retina1 ganglion cells was also carried out using DIC optics. We inner plexiform layer, respectively (Peichl and Wlssle, I98 1; Wiissle et measured the depth ofganglion cell dendrites with reference to the inner al., 198la,b). With one or two exceptions, all alpha and beta cells in border of the inner nuclear layer or the equator of ganglion cell bodies regions chosen for analysis could be confidently classified. For every using a linear encoder (Heidenhain M25; 0.1 pm resolution; accuracy, cell that we categorized as on-center or off-center, we first measured the 0.3 pm over 25 mm). The tip of this encoder rested on the stage. High- distance (depth) from the equator of the ganglion cell to the position of magnification plots of ganglion cells (e.g., Fig. 7) were made by over- secondary and tertiary dendrites in the inner plexiform layer. The ac- laying the video image of the retina onto a graphics program with curacy of these measurements was about 0.5 pm. Gamma cells could variable magnification (as shown in Fig. 110. The coordinates of gan- generally not be classified due to their poor dendritic labeling. Similarly, glion cells on these plots are accurate to within 10 pm. Low-power plots cells within 2 mm of the area centralis could not be identified reliably were made using a drawing attachment and a 4x planapochromatic due to poor dendritic filling. In classifying cells in the wildcat, we have objective. relied heavily on criteria developed in the domestic cat (Boycott and The classification of retinal ganglion cell types (on-center vs. off- Wlssle, 1974). center; alpha, beta, and gamma cell types) was carried out using a 100 x For the high-resolution sampling of ganglion cell densities in the area 1.25 NA planachromatic objective and DIC optics. On-center and off- centralis and along the horizontal meridian, the distance between ad-

Figure 1. Left and dorsal lateral views of the brain of Felis silvestris tartessiu, the Spanish wildcat (female S2). Sulcal patterns in this 3 1 gm brain conform to the type III pattern seen in about 18% of domestic cats (Fig. 7 in Otsuka and Hassler, 1962). The Journal of Neuroscience, January 1993, 13(l) 211

Table 1. Continued

P4 M, Protoc. Protoc. Parac. Statistical Length Width Height length Length Width height height Student’s t” resultsh 7.0-7.5 2.6-2.8 4.34.7 3.4-3.6 7.2-8.1 3.0-3.5 4.6-5.2 4.4-4.6 7.15 2.70 4.52 3.48 7.65 3.20 4.92 4.47 t = 5.88 p<.05atdf=3

8.30 3.65 5.20 4.55 9.35 3.95 5.60 4.75 7.90 3.35 5.10 3.80 8.45 3.75 5.30 4.65 8.10 3.50 5.15 4.18 8.90 3.85 5.45 4.70 reference reference 7.3-7.9 3.5-3.9 4.4-5.2 3.7-1.2 7.4-9.0 3.94.5 5.8-6. I 4.24.7 7.57 3.67 5.07 3.97 9.05 4.20 5.95 4.45 t = -0.04 no difference 6.2-8.2 2.9-3.5 4.4-5.2 3.54.0 7.4-9.0 3.2-3.9 4.7-5.7 3.2-4.4 7.10 3.26 4.83 3.71 8.26 3.54 5.25 4.03 1 = 7.11 p < .Ol atdf=3 7.49 3.56 5.54 4.08 8.74 3.80 - 4.28 t = 2.41 p < .05atdf= 11

7.96 3.71 5.57 4.41 9.48 4.11 4.98 3.9 t = 0.64 no difference

8.08 3.89 5.49 4.28 9.62 4.21 - - I = -3.30 p < .05atdf=3

jacent sample sites varied from 35 wrn in the area centralis to 200 pm and nuclear staining as described in detail in Williams and Rakic (1988a). in the far periphery. Sites were studied with a 100 x 1.25 NA planachro- Neuronal nuclei, both of principal neurons and intemeurons, were matic objective. Each site had an area of 1200 pm*. Sampling and counted in translaminar probes extending through layers A, A 1, C, C 1, counting procedures were unbiased (Brrendgaard and Gundersen, 1986) C2, and C3 (and the medial interlaminar nucleus in the most medial and no adjustments were made to avoid blood vessels or axon fascicles probes). Between nine and 11 probes were made through each nucleus or small tears in the whole-mount. For the purpose ofcounting, we used in regions representing upper, lower, central, and peripheral visual fields. the nucleolus as the criterion structure. Data were collected separately Each probe was made up of a series of adjacent counting boxes. The for labeled and unlabeled cells in the ganglion cell layer. We excluded depth of the counting box-the z-axis of the section-was optimized glial cells, characterized by their small bodies and minuscule nucleoli for the thickness of the individual section. We used the linear optical (Wong and Hughes, 1987a) from analysis. To determine peak densities encoder to define the tops and bottoms of counting boxes. The average of ganglion cells in the area centralis, we used video-enhanced DIC density of neurons in the counting boxes was calculated and multiplied optics to analyze sets of highly magnified fields (450 Frn2 at 6500 x). by the volume ofthe nucleus to obtain the total LGN neuron population. To calculate the peak density of ganglion cells we used the highest The volume of the LGN was determined by direct integration. To do average of six adjoining fields. so, the areas of eight or more sections through each nucleus were mea- Numbers of axons in the optic nerve. We prepared cross sections of sured and plotted against the rostraleye through the previous work (Williams and Chalupa, 1983; Williams et al., 1986). points, and the area under this curve-an accurate estimate of total Counts of axons were obtained in a systematically distributed sample LGN volume-was determined. Although this method is not as simple of fields (typically SO-loo), and our estimate of average axon packing as several alternatives described in Rosen and Harry (1990) it provides density was multiplied by the cross-sectional area of the nerve to obtain a useful graphic assessment of the shape of the nucleus, and it allowed an estimate of the total axon population. A change in sampling protocol us to detect and correct errors made during drawing and measuring the was required because of the large area of the wildcat optic nerves. The LGN cross sections. ultrathin sections of the wildcats were too large to fit onto single grids. The size distribution of neurons in the LGNs ofwildcats and domestic Consequently, several grids had to be photographed to sample the entire cats was studied in celloidin-embedded tissue. The cross-sectional areas cross section of the nerve. Electron micrographs covered fields of 467 of cell bodies were measured directly onscreen at 2000 x using a video- @ml. The sample area was determined with the aid ofa mesh calibration enhanced DIC overlay system (Williams and Rakic, 1988a) through grid photographed at the end of the sampling session. Axons in cats are layers A, Al, and the C layers. A 100 pm graticule divided into 2 rrn unbranched, and therefore the number of optic axons provides a good increments was used to calibrate all images. estimate of the retinal ganglion cell population (Chalupa et al., 1984; Lia et al., 1986). Results In the case of the fetal wildcat, 20-25 systematically distributed fields, each with an area of 93 pm*, were photographed and counted in each We have studiedtwo adult wildcats and one fetal wildcat. These optic nerve. Numbers of axons, growth cones, and necrotic fibers were animals were extremely difficult to obtain and there was no counted in each field, using the same criteria developed in Williams et immediate prospectfor increasingnumbers of cases.Therefore, al. (1986, 1991). our approach has been to analyze the material in detail, as if LGN cell analysis. We estimated the total population of neurons in the LGN in both frozen and celloidin-embedded material by means of we had uncovered a small number of rare and important fossils. direct three-dimensional counting using oil immersion objectives and In somecases, our analysisof the wildcats hasextended beyond DIC optics (Williams and Rakic, 1988b). The main advantage of three- data initially available for the domestic cat. This is true for the dimensional counting is that no correction factors-particularly, the analysesof photoreceptor distribution and dendritic differences often inaccurate Abercrombie correction-are required to estimate local cell density. Furthermore, this method is insensitive to variation in among on-center and off-center beta cells. Despite the small processing methods and section thickness. Neurons in the LGN were samplesize, the central resultsof this study-the marked differ- distinguished from glial cells and endothelial cells on the basis of size encesin populations of retinal ganglion cells and of geniculate 212 Williams et al. l Rapid Evolution of the Visual System

Figure 2. The photoreceptor mosaic and ganglion cell layers of the male wildcat SI. All images cover an area 78 x 100 pm. A and B are pairs of DIC images of precisely the same area of a retinal whole-mount, 0.5 mm above the area centralis close to the decussation line in temporal The Journal of Neuroscience, January 1993, 13(l) 213 neurons-do achieve statistical significance. With respect to 528 to 570 mm*, whereas the range among domestic cats is from other facets of this study, particularly those based on single 460 to 640 mm2 (Hughes, 1975; Chalupa et al., 1984; present animals, we have tried to be circumspect in our conclusions and results). Our huge domestic cat had retinal areas of 570 and 600 to point out interpretative caveats. In several instances, we pro- mm2. Linear retinal measurements are also equivalent: the dis- vide more complete documentation for the wildcats than for tance from the center of the area centralis to the center of the domestic cats. References to specific figures for domestic cats optic disk in the wildcat retinas is between 3.37 mm (S2 right) are made both in the text and in figure captions. and 3.55 (SI right), precisely in the same range as that reported in Nikara et al. (1968) Wassle et al. (1975) and Hughes (1975). Comparison of body weight and brain size of the two It is also likely that posterior nodal distances and retinal mag- species nification (about 213 wrn per degree near the area centralis) in The average body weight of F. silvestris tartessia males is about the two species are the same. 6.5 kg, and of females about 4 kg (Rodriguez de la Fuente, 1979). Our two wildcats were lean and weighed less than average: the Dlferences in rod and cone distributions male (S1) weighed 3.7 kg and the female (S2) weighed 3.3 kg. A quantitative analysis ofthe photoreceptor mosaic was possible Despite these modest body weights, the brains of the male and in both retinas from the male wildcat Sl. Data from this animal female wildcats weighed 37 gm and 3 1 gm, respectively. These demonstrate that densities of cone photoreceptors can reach values are far above the domestic cat average: 27.6 gm with an higher densities in wildcats than has ever been reported in do- SD of f 1.5 gm for male domestic cats, and 26.5 & 1.35 gm for mestic cats. The difference appears to be most marked in the females (Latimer, 1938). Even the brain weight of the small central retina, particularly at the center of the area centralis, female wildcat reproduced in Figure 1 is 3.3 SDS above that of where cone densities of up to 100,000/mm~ were encountered female domestic cats. Our huge 9 kg male domestic cat had a in this wildcat (Fig. 2). In comparison, the highest cone density brain weight of only 28.25 gm. This is 9 gm less than the brain we have yet encountered in domestic cats is 35,000~0,000/ weight of the less massive but equally tall male wildcat. In mmZ, a range that is itself about 25% higher than peak cone absolute terms, the difference in brain weight between the two densities reported in previous studies of this species (Steinberg adult wildcats and domestic cats is in the neighborhood of 20- et al., 1973, their Figs. 1, 2,6; Wassle and Riemann, 1978, their 30%. This difference conforms closely to that which we antic- Fig, 2). Cone densities are also high in the periphery of this ipated based on the allometric relationship between brain and wildcat’s retinas (Figs. 3, 4). However, the difference is not as body weights derived from a wide range of felids-brain mass great as in the area centralis. For example, along most of the = 0.23 (average body mass)0.6l (Davis, 1962; Radinsky, 1975; horizontal meridian, cone densities are only 20-30% higher than Pagel and Harvey, 1989). in the domestic cat at comparable eccentricities. Although not The pattern of gyri and sulci in wildcats conforms to the studied in the same detail, there also appears to be a comparative typical field pattern (Fig. 1; cf. Radinsky, 1975). The lateral, surplus of cones in dorsal and ventral retina of the wildcat in posterolateral, and suprasylvian gyri are arranged in a common comparison to the domestic cat. To obtain a rough estimate of pattern that Otsuka and Hassler (1962) refer to as type III. The the difference in the total cone population, we integrated the lateral and suprasylvian gyri are wider in the wildcats than is area under the curves in Figures 3 and 4. The ratio of these typical in domestic cats. For example, the lateral gyrus was areas is 1: 1.3 (domestic : wildcat). Therefore, along this axis, between 7.0 and 8.5 mm wide in the wildcats compared to domestic cats typically have 25% fewer cones than does wildcat typical values between 5.0 and 6.5 mm in domestic cats (in viva SI. values, Reinoso-Sdrez, 196 1). Although the breadth of the cra- Conversely, rod densities were much lower in the area cen- nium is greater in wildcat, the stereotaxic coordinates of bregma tralis of wildcat SI than has ever been measured in domestic and inion do not differ appreciably between species. The thick- cats (Steinberg et al., 1973; present results). Rod densities drop ness of bones of the cranial vault is much reduced in wildcats to well under 50,00O/mm* in the center of the area centralis of (e.g., the thickness of the temporal bone is merely 1.25 mm in SI (Fig. 2) but in domestic cats we have been unable to locate wildcat vs. 2.6 mm in domestic cats). In this case, the thick any fields in the area centralis of any animal in which rod den- skulls of domestic cats do house smaller brains. sities fall to less than 200,000/mm2. Along most of the hori- zontal streak of the wildcat, rod densities average between Analysis of retinal structure 300,000/mmz and 400,000/mm2. This is on the low side of the Constancy of retinal dimensions range we have observed in the horizontal streak of domestic There is complete overlap in the area of wildcat and domestic cats. In both species, rod densities range up to 550,000 in the cat retinas. The range among the four wildcat retinas is from dorsal periphery. To estimate the difference in the rod popu- t hemiretina (ipsilateral to the HRP injections). The two micrographs were taken at focal planes separated vertically by 74 pm. A, Micrograph at the level of the photoreceptor inner segments. Large cone inner segments stand out clearly among the small rod inner segments. A single cone is marked by a small drrowhead. This image can be compared to those reproduced in Figure 2 of Steinberg et al. (1973) and Figure 2 of Wissle and Riemann (1978). B. Micronrauh taken at the level of the ganglion cells. Most of these cells are labeled with the dark HRP reaction product, but at least one &abkled cell with -a large nucleolus is also a gan&on cell (arrowhead), presumably one that has a crossed projection: C and D, Comparable through-focus pairs of micrographs taken in the area centralis. Note that the cones in area centralis, one marked by an arrowhead, are somewhat smaller than in A and are much more densely packed (densities range from 50,000 to 100,000). Smaller rods are also scattered throughout the area centralis. We have not been able to find regions of such high cone density in domestic cats, nor have other investigators (see Steinberg et al., 1973, their Fig. 2). In D, only a single labeled ganglion cell is present in this field just nasal to the decussation line in the retina ipsilateral to the central HRP injections. Only 49 other ganglion cell bodies are visible in this focal plane. However, ganglion cells are stacked two or three cells deep here, and this small field actually contains 1 12 ganglion cells. This is equivalent to a density of 14,400 cells/mm2. Williams et al. l Rapid Evolution of the Visual System

Wildcat Photoreceptor Distribution Profile 90,000 T Slbsi HRF

:y Nasal

Figure 3. Gradients in wildcat cone and rod densities through the area cen- “E tralis and along the horizontal axis (male E 300,000 wildcat S1). The vertical axes in this z figure and Figure 4 (domestic cat) are (1 identical except that the y-axis is inter- p 200,000 rupted above 30,000 mm*. The gray bars at the far right and left of these 43 plots represent eccentricities at which we could not make accurate measure- ments because of excessive retinal pig- Temporal Nasal ment. A sketch of the ipsilateral HRP- : ! ! : 1 : : ! :I ( : : : ; ! ) ! ! ; : ! ! 1; ; : [ : 1 labeled retina in the upper right shows 10 the horizontal line along which data 5 0 5 10 15 20 were collected. T, temporal; N, nasal. distance from area centralis (mm) lation, we again integrated the area under the curves (Figs. 3, Stone et al., 1982; Chalupa et al., 1984; Wong and Hughes, 4). The ratio is 1.2: 1 (domestic : wildcat). 1987b). It is probable that the reduction in cone and ganglion cell densities in the area centralis of the domestic lineage has Marked dlferences in ganglion cell numbers and distribution been matched by a reduction in sampling resolution of both cell Total ganglion cell densities are substantially higher in all retinas arrays, as well as in photopic acuity. Ganglion cell densities in of both wildcats than in domestic cats, both in the area centralis the periphery are also greater in both wildcats than in domestic and across the retinal periphery (Fig. 5). For example, in the cats. However, the quantitative difference is less marked, av- area centralis, ganglion cell densities are 50% to nearly 100% eraging 15-25% (Fig. 5). For example, in the field of cells de- higher in the wildcats than in domestic cats. In particular, the picted in Figure 7, the density of HRP-labeled ganglion cells is peak density of HRP-labeled ganglion cells in the right retina 300/mm* at a mean eccentricity of 9.4 mm. At this eccentricity of SI (contralateral to the injected hemisphere) is 15,100 cells/ in the domestic cat, ganglion cell densities are typically -2OO/ mm* at a location 40 pm nasal to our estimate of the position mm* (Figs. 5, bottom; 8; see also Stone, 1978; Chalupa et al., of the center of the decussation line (Fig. 6A,B). At the decus- 1984). sation line itself, the density of labeled and unlabeled cells in Decussation pattern does not d$?&. The pattern of decussation Sl is 10,100 and 5,900 cells/mm*, respectively, giving a peak of ganglion cell axons was examined in wildcat Sl by retrograde density of about 16,00O/mm*. In wildcat S2, a case in which labeling from one hemisphere (Fig. 6AJ). The characteristics both retinas were stained with toluidine blue, the peak density of the decussation pattern and the relative sizes of the temporal of cells with large nucleoli in the area centralis is 18,400 in the and nasal retinal components appear indistinguishable at a qual- left retina (Fig. 60) and 20,900 in the right retina. In compar- itative level from those noted in domestic cats by previous ison, the highest packing density of HRP-labeled retinal gan- investigators (Cooper and Pettigrew, 1979; Jacobs et al., 1984, glion cells we have found in the center of the area centralis of their Fig. 1). a domestic cat is 10,500/mm2 (Fig. 5, bottom). This value The horizontal streak. We were struck by the prominence of matches peak ganglion cell densities reported in numerous pre- the horizontal streak in the wildcat retinas. However, our quan- vious studies of domestic cat retina (Hughes, 1975; Stone, 1978; titative analysis revealed that the increment in cell density in The Journal of Neuroscience, January 1993, 73(l) 215

Domestic Cat Photoreceptor Distribution Profile

“E E 300,000 az p 200,000 g 100,000 -- Figure 4. Gradients in domestic cat cone and rod densities through the area Temporal Nasal centralis and along the horizontal axis 0 !111111~:I:I:II:I!I(!l!!l:I::I (domestic cat 08). Faint gray curves in 10 5 0 5 10 15 20 both A and B are taken from data in Steinberg et al. (1973). T, temporal; N, distance from area centralis (mm) nasal. the wildcat streak is of the samemagnitude as that seenelse- domestic cats; cf. Wassle et al., 1975; Stone, 1978; Kirby and where in peripheral retina (Fig. 5). For example, the density of Chalupa, 1986). We counted 410, 425, and 428 HRP-labeled labeledretinal ganglion cells in the mid-nasal periphery of the alpha cells in a 3-mm-high x 2-mm-wide region located just visual streak (6-12 mm eccentricity) averages 1000 cells/mm2 nasalto the areacentralis in three domestic cats. In comparison, in wildcat SZ and about 600-700 cells/mm* in domestic cats we counted 411 HRP-labeled alpha cells in the same area in (Rowe and Stone, 1976; Chalupa et al., 1984). wildcat SI. Similarly, we identified a total of 14 12 Nissl-stained Given the lower densitiesof ganglion cells in the domestic alpha cells in the 36 mm* region centered on the area centralis cat, most prominently around area centralis, one might expect in wildcat S2 (Fig. 9) while Whsle et al. (1975, their Figs. 6A, a compensatory increasein the spreadofdendritic fieldsin order 8A) identified 1484 alpha cells in an equal area in a domestic to conserve field overlap (Whsle et al., 1981 b). To answerthis cat-a difference of merely 5%. Densities of alpha cells are also question unequivocally would require more complete dendritic remarkably close in the mid and far periphery of both species. filling than we have been able to achieve in either beta or gamma For instance, the density of HRP-labeled alpha cells 9-10 mm cell classes.However, the marked speciesdifferences in ganglion above the areacentralis in dorsal retina just nasalto the decussa- cell density at the area centralis and the sharp peak in cone tion line is 14/mm2in wildcat SI and 1 l/mm* in wildcat S2. density in wildcat SI suggestthat photopic visual acuity is high- In domestic cats, values in this region range between 10 and er in wildcat than domestic cat. 13/mm2(Wassle et al., 1975, their Fig. 6A; Stone, 1978, his Fig. Conservation of alpha cell number and distribution. While the 5). Thus, it seemshighly probable that the total alpha cell pop- total ganglion cell density is higher in wildcats than has ever ulation in wildcats is in the samerange as that for the domestic been reported in the domestic cat, this is not true of one par- cat-5000 to 7000 (Wong and Hughes, 1987a). ticular ganglion cell class: densitiesof alpha cells are almost No decline in alpha cell density in the wildcats’ area centralis. precisely the samein both species(Fig. 9). For example, alpha In wildcat Sl, the pattern of HRP-labeled cells allowed us to cell density in the area centralis is between 120 and 150 cells/ define the location of the center of the area centralis with a mm* in both wildcats, whereasin domesticcats, alpha cell den- precisionof f 25 pm, independentof the alpha cell distribution sity is between 130 and 200 cells/mm2 (our data from three itself (Fig. 6A,B). This made it possibleto assesswhether there 216 Williams et al. - Rapid Evolution of the Visual System

t a,a, ,...,7s-1* ...... L

g Domestic Cat DC75 ...... ~.~ &p$ (- ...... c

2,500 Figure 5. Central-to-peripheral gra- dient of ganglion cell density along the horizontal axis in wildcat SI (top) and domestic cat DC15 (bottom). Qualita- tively, the gradients are the same in both species, but densities are higher, most prominently around the area centralis of the wildcats. To obtain the plot in 250 ...... 4 ...... + ...... I”““” ,...... ,,.,._ the top graph,numbers of HRP-labeled Nasal : retinal ganglion cells were counted and / 0) summed in both retinas ipsilateral and 100 contralateral to HRP injections (see Fig. 6). We sampled data in a completely 0 5 0 5 10 15 20 unbiased manner and include fields tra- versed by large blood vessels. This ac- counts for some ofthe variability among data points. distance from area centralis (mm) is a local decline in alpha cell density in the wildcat area cen- which the assignmentof cell types is unambiguousare similar tralis-an unresolved issuethat has arisen in the domestic cat in the two species.However, given the finding that the alpha retina (WBssleet al., 1975; Stone, 1978; Mountcastle, 1980, p cell population is the same in both species,but that the total 539). In the wildcat, the density of alpha cells within a radius population of ganglion cells is higher in wildcats, it is clear that of 100 Frn of the center is 127/mmz, and in two slightly more the overall proportion of alpha cells must be slightly lower in peripheral annuli, densitiesare 121 and 132 alpha cells/mm2. wildcats than is typical in domestic cats (cf. Stone, 1978; Wong Nor was a central decline in alpha cell density noted in S2 (Fig. and Hughes, 1987a). Furthermore, at the area centralis of the 9). Similar analysesof HRP-labeled retinas from three domestic wildcat, a region in which the total ganglion cell density is par- cats also did not demonstrate any appreciable decline in alpha ticularly high, alpha cells make up a smaller percentage(< 1%) cell density at the precisecenter of the area centralis. of the cell population than in domestic cats (2-3%). Relative abundance of alpha, beta, and gamma ganglion cells As has also been noted in the domestic cat (Wgssleet al., are similar. The numbers of the alpha, beta, and gamma cell 198 1b; see also Fig. 8), there are slightly more off-center beta types were studied in the dorsal periphery of wildcat Sl. Here cells than on-center beta cells (26.9% off-center, 22.7% on-cen- the proportions of HRP-labeled ganglion cell types are similar ter, 0.6% unknown) in the wildcat dorsal periphery (Fig. 7). to those in domestic cats (Figs. 7, 8). Beta cells make up 50.2% Correspondingpercentages for the domestic cat are 26.2% off- of the local ganglion cell population (155 of 309 cells), gamma center and 22.7% on-center cells (Fig. 8). cells make up 45.3%, and alpha cells make up the remainder Ganglion cell size is the same.The higher averagecell density (I 4 of 309 cells in an area of 1.08 mm’). In the domestic cat, in the wildcat retina could be associatedwith a reduction in beta cells make up 48.9%, gamma cells make up 45.3%, and mean cell body size. Such changeshave been found following alpha cells make up the remainder (13 of 225 cells in an area experimental manipulations in domestic cats (Kirby and Chalu- of 1.08 mm*), Thus, ratios of cell types in a region of retina in pa, 1986). We analyzed the sizesof all the cells in Figures 7 and The Journal of Neuroscience, January 1993, 13(l) 217

Figure 6. The area centralis of the wildcat retina and the precision of the line of decussation. A, The central region of the retina contralateral to hemispheric HRP injections in the male wildcat SI. B, The ipsilateral left retina from the same animal. The regions shown in A and B are 1.08 mm x 1.43 mm. For alignment purposes, the left retina in B has been oriented so that dorsal is down. In the wildcat, as in the domestic cat, there are numerous cells in temporal hemiretina that have crossed projections (left side ofA), but there are very few cells in the nasal hemiretina that have uncrossed projections (right side oJB). Comparable figures of the line of decussation in domestic cat are provided in Jacobs et al. (1984, their Fig. 1). C, Higher-magnification micrograph of the center of the area centralis. Note the relatively abrupt change in patterns of decussation as revealed by the degree of overlap of labeled and unlabeled retinal ganglion cells. Arrowheads in A and C point out a common landmark-a large alpha cell. Magnification in C is 342 x . D, The entire area centralis in a Nissl-stained preparation from wildcat S2. The large darkly stained cells are alpha cell bodies. Alpha cells in the center of the area centralis do not stain as darkly, in part because the ganglion cell layer is comparatively thick. The region of the wildcat area centralis in which ganglion cell densities are above 10,000/mm2 is small, corresponding approximately to a circle with a radius of 150 pm (see Fig. 9). In this zone, the ganglion cell layer is typically two cells thick. Magnification in D is 72 x , and the field is 1.36 mm x 1.25 mm in horizontal and vertical dimensions. 218 Williams et al. + Rapid Evolution of the Visual System

0 Figure 7. Retinal ganglion cell mosaic cc in the wildcat Sl. This field ofcells cov- ers an area ofabout 1 mm* in the dorsal 00 part of the retina. Coordinates are in- 9.0 dicated on the axes. Large off-center al- 0 pha cells are represented by black ir- 0 regu/urshapes. On-center alpha cells are similar, but white. On- and off-center g8.9 beta cells are represented by the large circles-black for off-center, whitefor on-center, and gray for two beta cells of unknown subtype. The small circles represent the heterogeneous gamma cell class. Only a few of these cells could be categorized as on-center or off-center. distance from line of decussation

8. In wildcat SI, the averagecross-sectional area of 309 ganglion in another domestic cat, DC13. It is also the casethat HRP- cells at a location 9 mm above the area centralis is 357 pm* labeled alpha cells in the wildcat SI are smaller than in the (equivalent to an averagediameter of 2 1.3 pm). In the domestic domestic cats (Figs. 7, 8, 10). Given the small sample size, it cat, the average size of the 230 ganglion cells in a comparable seemsprudent simply to concludethat data on cell size in wild- region, plotted in Figure 8, is almost precisely the same-359 cat SI fall within the range seenin domestic cats. km*. This equality in ganglion cell size is remarkable given the Unsuspecteddendritic d@erencesbetween on- and @center sizable difference in the local density ofganglion cells-286 cells/ beta cells are conserved.The dendritic morphologiesof on- and mm2 in the wildcat and 208/mm* in the domestic cat. off-center beta cells in the wildcat differ consistently, both in More subtle differences in cell size between speciesare, in the depth of their arborization (the standardcriterion) and also general, not masked by averaging acrosscell classes(Fig. 10). in the shapeof their proximal dendritic arbors (Figs. 11, 12). For example, the average beta cell diameter in SI and the do- On-center cells almostinvariably have three to five thin tapering mestic cat DC13 are both 23 pm. Similarly, the average di- and radiating primary dendrites (Figs. 11B,E; 12) that spread ametersof gamma cells in Sl and DC12 are both about 15 pm. out directly from the cell body and arborize neatly in the inner However, cell size is quite variable among domestic cats, even half of the inner plexiform layer at a distance of 6-7 pm from at equivalent retinal coordinates. For instance, in domestic cat the equator of the cell body. In contrast, off-center cells almost DC12, the mean size of HRP-labeled retinal ganglion cells 9- invariably have one or two thick primary dendritesthat ascend 10 mm above the area centralis was 472 pm2 versus 359 grn2 sharply and then arborize in the outer third of the inner plexi- The Journal of Neuroscience, January 1993, 13(l) 219

ti+ t-3 0 F 9.60

i.

ii o. Iii 5 0 9.4 (

Figure 8. Retinal ganglion cell mosaic in the domestic cat at a comparable lo- cation to that shown in Figure 7 for the wildcat. All conventions are as in Fig- 1.3 1.5 1.7 ure 7. The dendritic arbors of alpha cells 0 0 l 0 00 0 0 have been drawn somewhat more com- distance from line of dec%ation (mm) pletely in this plot than in that shown for the wildcat.

form layer (Figs. 1 lA,D,F; 12). Off-center arbors also have a reexamination of retinas from several domestic cats revealed more complex appearance than those of on-center beta cells, precisely the same dendritic differences first noted in wildcat and the branch points of off-center arbors are often flared out- Sl. ward (Fig. 11). Off-center beta cell dendrites are almost invari- ably more curved (spraylike) than those of on-center cells and are broader and more heavily labeled with HRP. Off-center Analysis of the optic nerve arbors are also most commonly disposed asymmetrically with Marked species d$erences in number of optic axons in adults respect to the cell body, whereas on-center cell dendrites radiate The cross-sectionalareas of the wildcats’ optic nerveswere sub- symmetrically. In wildcat Sl, on- and off-center cells can be stantially greater than those of domestic cats: 2.95 mm2 in SI reliably identified using these characteristics without examining and 3.24 mm2 in S2. This compares to a mean of about 2.0 the depth of the dendritic arbor. These types of dendritic dif- mm* in domestic cats (Williams et al., 1986). The packing den- ferences between on- and off-center beta cells have not been sity of fibers, however, is in the same range-8.8/100 firn2 in reported previously in the domestic cat, and because these cells Sl, 7.22/100 Nrnz in S2 (Fig. 13A) and -8.0/100 wrn’ in do- have been studied in great detail, we initially assumed that this mestic cats (Williams et al., 1983). Given these findings, it is distinction must be unique to the wildcat. However, careful not surprising that the total population of axons is much higher 220 Williams et al. * Rapid Evolution of the Visual System

Wildcat Sl

6 12 16 20 24 26 32 36 : 40 12 T 10

$6 E Domestic Cat DC72 g 6 & Q 4

2

3 2 1 0 1 2 3mm 0 Figure 9. Alpha cell distribution in and around the area centralis of wildcat S2. The center of the area centralis, a region is which total l2 T ganglion cell densities in wildcat are above 10,000/mm2 and in which cone densities are elevated, is encircled. The region reproduced in Figure 60 is outlined by the solid-line rectangle. Comparable data for alpha Domestic Cat DC13 cell density in central retina of the domestic cat are provided in Figure &I of Wassle et al. (1975).

in wildcats than in domestic cats: 260,000 f 6300 in SI and 234,000 + 4500 in S2 versus 150,000-165,000 in domestic cats 2 (Chalupa et al., 1984; Williams et al., 1986). This difference between speciesis significant (t = 6.7, p < 0.05, two-tailed test; 0 domestic cat data taken from Chalupa et al., 1984; Williams et al., 1986). ganglion cell diameter (pm)

Axon populations in the nerves of the fetal wildcat Figure 10. Ganglion cell size in a wildcat and two domestic cats at Gestation in the wildcat is of the sameduration as that of the sites 9 mm above the area centralis in the nasal hemiretina. HRP-labeled cells were measured. Note the substantial variation in the two lower domestic cat (63 + 3 d); weight at birth is about the same, as histograms taken from two domestic cats. is the ageat eye opening(Lindemann and Rieck, 1953;Hemmer, 1976; Rodriguez de la Fuente, 1979). We therefore estimated the age of the fetal wildcat using data on the tempo of devel- high-an average of 800/ 100 pm2 (Fig. 13B). This is very close opment in the domestic cat (Williams and Chalupa, 1982; Shatz, to the value of 7351100 pm2 in the domestic cat fetus at E39. 1983; Williams et al., 1986). On the basisof parameterssuch The right and left optic nervesofthe wildcat fetuscontain 670,000 as crown-to-rump length, body weight, eye mass, and retinal f 30,000 and 540,000 + 30,000 axons, respectively. The dif- surfacearea, we conclude that the developmental stageof this ference between right and left nerves is high and could be due wildcat fetus correspondsvery closely to that of fetal domestic to undercounting in the left nerve. However, there was no par- cats between embryonic day 38 (E38) and E40 (Table 2). ticular technical difficulty in counting either nerve, and it seems The density of axons in the fetal wildcat nerves is extremely as likely that the difference reflectssmall variation in the tempo

Figure 11. On-center and off-center ganglion cells in the wildcat. A, Three off-center beta cells (arrowheads) have dendrites located in the inner plexiform layer in a focal plane 12 pm away from the location of their cell body centers. B, Typical thin dendrites of two adjacent on-center beta cells located 5 Nrn distant from the center of their cell bodies. C, A single field photographed at two focal planes in the inner plexiform layer. The upper halfof the field is filled by the large dendrites of an off-center beta cell at a depth of 15 pm. The lower halfshows part of the arbor of an on- center beta cell at a depth of 8 pm. This field is located 2.5 mm dorsal to area centralis. D. Typical asymmetrical orientation of off-beta cell dendrites. Note the large caliber of the dendritic trunk. E, Large on-center alpha cell, six beta cells, and three gamma cells. The identity of each cell in this field is depicted schematically in Figure 7 at a location 9.3 mm dorsal to area centralis and 1.75 mm nasal to the decussation line. -7, Video image and drawing of an off-beta cell (see Fig. 12) viewed on the monitor using video overlay. The drawing of the cell has been offset slightly above the black-and-white image of the HRP-labeled cell. It is possible to plot large fields of cells at high magnification using this system. Magnification: A, B, D, and E, 530 x ; C, 1325 x ; F, 800 x (1600 x on the monitor). A

n 222 Williams et al. l Rapid Evolution of the Visual System

Table 2. Comparison of fetal wildcat and fetal domestic cat

Domestic cats, Fetal wildcat E38/E39 Crown-rump length 65 mm 66-71 mm Weight 19.4 gm 2&22 gm Brain weight 570 mg 560 mg Eye weight 40 mg 35-48 mg Area of retinas 30 mm2 20-25 mm*

of development in the two eyes (seeWilliams et al., 1991). In ON any case, the estimates from this animal are remarkably close to thoseof two E39 domestic cat littermates we have previously 0 studied-698,000 + 20,000 and 557,000 f 28,000 (Williams et al., 1986). The percentagesof growth conesand necrotic fibers in the wildcat fetal nerves are also very closeto those found in domestic cats at E39 (Table 3). These two parametersare sen- IPL ceiling = 12 pm sitive indicators of the rate of change in the fiber population during development. The low density of both growth conesand 6.5 I : i I : I 10 pmldiv necrotic fibers (Table 3) indicates that the population of fibers counted in these fetal wildcat nerves is at or very close to the Figure 12. Typical morphology and dendritic depth of on- and off- ontogenetic peak. center beta cells in the inner plexiform layer of wildcat SJ. The depths of different dendritic segments were measured from the equator of the Analysis of the dorsal lateral geniculate nucleus cell body. The interface of inner plexiform and inner nuclear layers was typically 12 f 1 cm from the middle of the ganglion cell layer in the The volume of the LGN is less in domestic cat mid periphery of this retina. There is no appreciable overlap in the The volume of the LGN calculated from both frozen and cel- depth of secondary and tertiary dendritic branches of on- and off-center loidin sectionsis substantially greater in the wildcat than in the beta cells. A video-overlay image of the off-center beta cell in the upper domestic cat (Table 4). For example, the left LGN volume of left quadrant is shown in Figure 11F. IPL, inner plexiform layer. domestic cat Dl is 28.1 mm), whereascorresponding values for the two seriesof frozen sectionsfrom the wildcats are 38.1 and

Figure 13. Electron micrographs of optic nerve fibers in wildcat. A, Micrograph of a field in the left nerve of S2 at the same magnification (11,000 x ) used to count axons. Due to substantial fiber and myelin distortion, fiber diameter was not measured. B, Fibers in the left optic nerve of a fetal wildcat. This micrograph is also reproduced at the magnification at which the analysis was performed (15,000 x ). Arrowheads mark four growth cones and their shanks and, in the lower Iej, one large necrotic fiber. Comparable micrographs from a domestic cat fetus arc reproduced in Figure 21 of Williams et al. (1986). Scale bar, 1 Wm. The Journal of Neuroscience, January 1993, 13(l) 223

Table 3. Analysis of optic axons in fetal wildcat and fetal domestic cats

Fetal wildcat Left Right Domestic, E390 Domestic, E39” Axon population 536,000 f 29,000 668,000 f 29,000 557,000 f 28.000 698,000 + 20,000 Area of nerve (pm2) 75,000 73,000 74,400 95,700 Axon density (100 Frn ?) 714 911 749 729 Necrotic fibers (%) 0.07 0.16 0.21 0.26 Growth cones (%) 0.17 0.15 0.20 0.20 ” Data from E39 domestic cat littermates from Williams et al. (1986, their Table I).

35.5 mm’. After embedding the other hemisphereof the same cell size between domestic cats and the wildcats (Table 5). In domestic cat in celloidin, the LGN volume is reduced by pro- both species,the smallestinterneurons range in size from 50 to cessingto merely 10.4 mm’. Corresponding LGN volumes for 80 Frn*, whereasprincipal neurons range from 90 km’ to more the two wildcat hemispheresfixed, embedded,and cut in pre- than 400 Frn2. Even a comparison of the size distribution in cisely the same manner are 17.9 and 18.9 mmz. A reasonable wildcat and domestic cat caseswith large differencesin LGN estimate basedon our own material and that kindly lent us by volume (18.9 mmz vs. 10.4 mm3) demonstratesa remarkable Drs. Frost and Elberger (Table 4) is that the wildcat LGN has similarity in the sizesof neurons in the two species(Fig. 14). a volume that is roughly half again as large as that of the do- In comparison with the domestic cat, it doesnot appear that mesticcat LGN. The relative volumes of major componentsof any particular layer differs either in volume or in cell number the LGN are similar in domestic cats and wildcats (Table 4). In more or lessthan the nucleusas a whole (Table 4). It is of course both species,the A layer occupies about 37% of the nucleus, possible that the relative abundance of classesof LGN neu- whereasthe Al layer occupies 28%. rons-particularly classes1 and 2 of Guillery (1966)-differs between wildcat and domestic cat. Unfortunately, these cell Marked speciesd@erences in total neuron number classescannot be distinguishedas easily in the LGN as they can As the foregoing analysismight lead one to suspect,the neuron in the retina. population of the wildcat LGN is much higher than that of the domestic cat LGN (Table 4, right column). The populations of LGN neurons in the two wildcats were 766,000 and 754,700 Discussion (averages of left and right nuclei for Sl and S2, respectively). The descentof the domestic cat In comparison, the averagepopulation in five domestic cats was The speciesF. silvestris dates back to the middle Pleistocene only 5 10,000 + 26,000. These estimates include interneurons (Holsteinian interglacial period; Kurten, 1965b, 1971). During and principal neurons, and our values for the domestic cat are the late Pleistocene, the ancestral population of large wildcats extremely close to those calculated by Madarasz et al. (1978; gave way to the smaller European and North African wildcat their estimate: 555,000 neurons). If we make the reasonable subspecies,F. silvestris silvestris and F. silvestris lybica. These assumption that LGN cell populations are distributed normally subspeciesprovided the stock from which domestic cats trace in both species,then this speciesdifference is highly significant their descentover 3000 years ago (Zeuner, 1963; Wayne et al., (t = 15.34, p < 0.05, two-tailed test). 1989). Although details of this phylogenetic reconstruction may Cell size in the LGN. Neurons in the two speciesare very be modified as more material is recovered, there can be little closely matched in size. Sets of between 75 and 200 neurons doubt that among extant wildcats, the Spanish subspeciesis were measuredin layers A, A 1, and the C layers, in two or three morphologically closestto the Pleistocenestem population. sectionsfrom each of six animals. We found no difference in Dwarjing. There have been at least two distinct episodesof

Table 4. LGN volume and neuron populations

Meth- Volume Layer Neuron number O& (mm’) Ah Al c-C3 +- SEM Wildcat Sl right C 18.9 7.0 (37%) 5.2 (28%) 4.5 (24%) 774,300 * 53,000 Wildcat Sl left F 38.1 75 1,800 f 33,000 Wildcat S2 right C 17.9 739,700 f 132,000 Wildcat S2 left F 35.5 769,700 +- 40,000 Domestic Dl left F 28.1 Domestic DI right C 10.4 4.0 (38) 3.3 (32) 2.5 (24) 540,900 + 28,000 Domestic Juan C 16.0 5.9 (37) 3.9 (24) 3.8 (24) 475,200 f 35,000 Domestic Storm C 15.6 5.1 (33) 4.5 (29) 4.0 (26) 456,000 f 75,000 Domestic Pulomino C 21.7 5 14,200 + 62,000 Domestic MCI F 24.7 527,200 + 44,000 * C, celloidin-embedded material; F, material cut frozen. fi The percentage contribution of each layer to the total LGN volume is provided in parentheses. 224 Williams et al. - Rapid Evolution of the Visual System

Table 5. LGN neuron size the exceptional work of Menner, 1939; Rohrs, 1955; Ebinger and Lohmer, 1987). One problem facing this type of analysisis Layer that changesin brain structure associatedwith altered body size A Al c-c3 must be dissociatedfrom those effects due only to domestica- tion. In our cellular analysisof retina and LGN of cats, we have Wildcat SI 168 170 126 also been unable to dissociatethese effects. Although we have Wildcat S2 134 143 131 shown that the difference in average body size of the Spanish Domestic Dl 166 160 103 wildcat subspeciesand the domestic cat is sufficient to account Domestic Juan 174 184 164 for the overall reduction in brain massin the domestic lineage Domestic Palomino 164 170 144 (seeResults), it is probable that specificselective pressures,par- Domestic MCI 153 159 100 ticularly those associatedwith cohabitation with humans,have Cross-sectional areas (pm’) were measured at the focal plane in which this value also generated specific changesin brain structure. A compre- reaches a maximum. The SEM for each value is between 3 and 9 wm*. All material, hensive comparative analysisof other componentsof the CNS except that from case MCI, was embedded in celloidin. In case MCI, the brain was cut frozen. There is a surmising degree of variation among the mean size of and of different wildcat subspecieswould be particularly useful cells in the C layers of domestic cats (range, 100-164). This variation appears to in sorting the relative importance of size reduction and domes- be independent of cell size in the magnocellular layers. tication in driving brain evolution. It would be especially in- terestingto study the brain ofthe smallerNorth African wildcat, selection in the lineage that has led to the domestic cat. The F. silvestris lybica. first episode has been associatedwith a rapid decline in body size. This reduction occurred at the closeofthe Pleistocene,long Brain to body size scaling: a cellular analysisof allometry before domestication (Kurten, 1965a,b; Table 1). Such dwarfing In this study we have focused on quantitative evolutionary is by no meansexceptional; extremely rapid reduction in body changeswithin a small part of the CNS. Our data provide a size has been widespread among the Pleistocene mammalian cellular perspectiveon allometric changein the size of the brain fauna of Eurasia, North America, and Australia (Kurten, 1959; and body. This cellular level of analysiscomplements previous Marschall and Corrunccini, 1978; Stanley, 1979; Raff and Kauf- work on brain-body allometry, which has generally focusedon man, 1983). With the exception of the Spanishwildcat subspe- such global factors as ecological niche, taxonomic level, meta- cies, the reduction in body size in the F. silvestris lineagehas bolic load, and developmental timing (Eisenbergand Wilson, proceeded at a particularly rapid rate-about 30 darwins over 1978; Lande, 1979; Armstrong, 1982; Martin and Harvey, 1985, the last 20,000 years (Km-ten, 1958, 1959, 1965b, his Fig. 10; assessedin Mann et al., 1988; Preuss,1992). one darwin is equivalent to an e-fold change over 1 million The averagesize difference between the Spanishwildcat and years). On the basisof our data, and conforming to allometric domestic cat is approximately twofold-6-7 kg for the wildcat tendencies,the tempo of change in the visual system appears versus 2.5-3.5 kg for the domestic cat. The 25-30% difference to have proceededat a reduced, but still very rapid rate (about in averagebrain weight-31-37 gm for the wildcats versus25- 15 darwins). As in other lineages,the reduction may have been 29 gm for the domestic cat-is almost preciselywhat one would linked with postglacialclimatic changeand human competition predict from the allometric relationship calculated for felids (Kurten, 1965a, 1988; Van Valen, 1969; Martin and Klein, (Davis, 1962). We note that the two wildcats we studiedweighed 1984). lessthan is typical, presumablydue to their prolongedcaptivity. Domestication. The second episode of intense selection has However, this doesnot compromise our observation that their been associatedwith domestication. This processhas undoubt- brains are much larger than those of domestic cats, and that the edly causedchanges in brain, behavior, and reproduction (Dar- weightsof their brains are consistentwith allometric predictions win, 1890; Zeuner, 1963). However, the effectsof domestication basedon the average body weight of the Spanishwildcat. on brain structure have not yet been studied in detail (but see We can now ask and answer more interesting questions:how

80

UY70 s 'i 60

F 50 5 z 40 Figure 14. Size distribution of neu- f 30 l-lI ronal cell bodies in the LGN of wildcat Sl, and a domestic cat. In both cases, = 20 1000 cells were measuredin sample ,’ regions extending through all laminae. 10 I We usedcelloidin-embedded tissue for Ii this analysis.These area1measure- o...... ments of cell size greatly underestimate 150 200 250 the in vivo valuesdue to tissueshrink- age (Weber and Kalil, 1983). cross-sectional area (pm’) The Journal of Neuroscience, January 1993. 13(l) 225 has the allometric reduction in brain weight in the cat lineage been accomplished? What are the cellular correlates or mech- A subtle d@rence between on- and @center beta cell anisms? Our results reveal that in the primary visual system, dendrites is conserved in both Felis species the decrease in brain mass has been brought about strictly by We initially thought that the marked dendritic differences noted reductions in cell number. In particular, we have shown that between on- and off-center beta cells (symmetric vs. asymmetric populations of retinal ganglion cells and neurons in the LGN and thick vs. thin primary dendrites) were unique to the wildcat are reduced 30-35%. This reduction matches the reduction in (Figs. 11, 12)-that we had finally detected a significant quali- total brain weight. There is no evidence for any change in neuron tative difference. However, careful inspection of domestic cat size or in the amount of neuropil. If one is willing to generalize retinas revealed precisely the same structural differences be- our findings in retina and the LGN to the entire CNS, then it tween on- and off-center beta cells. Although we are the first to follows that the total population of neurons is about 30% less comment on this dichotomy between on- and off-center beta in the domestic cat than in the wildcat. cells, some of these differences can be discerned in figures of These data suggest that rapid evolutionary modulation in previous reports on domestic cat ganglion cells (e.g., Wassle et brain size is brought about primarily by changes in cell number. al., 198 lb; Rowe and Dreher, 1982; Stanford, 1987). These Wikler et al. (1989) have also noted that the evolution of the marked structural differences in dendrites suggest that a func- retina in the Cricetidae family has primarily involved change tional reanalysis such as that of Koch et al. (1982) would be in cell number. In contrast, long-term brain evolution encom- revealing, and that these two subtypes of neurons may have passing many millions of years of divergence has often been unsuspected electrotonic differences. From an evolutionary per- associated with substantial changes in cell type and cell size, spective, it is also noteworthy that relatively subtle, previously dendritic and electrophysiological properties, and the number undetected, dendritic differences have been conserved in this of cytoarchitectonic areas (Holloway, 1968; Purves et al., 1986; lineage despite large differences in numbers of these cells. An Kaas, 1987; Purves, 1988; Bekker and Stevens, 1990; Reiner, analysis of on- and off-center cells in other mammals would be 1991). Such changes must usually involve more far-reaching of interest to assess the taxonomic breadth of this dichotomy. genetic and epigenetic reprogramming than those associated with fluctuations in cell number alone. Developmental mechanisms and brain evolution Allometric exceptions. We found several interesting excep- An intriguing question is raised by these findings: at what stage tions in which numbers of cells appear not to have changed as of development and in what manner do the great differences one would predict from the allometric relationship between brain between the brains of wild and domestic species arise? As men- and body size. First, the population of rods in the domestic cat tioned in Results, the length of gestation in wild and domestic equals, and may even exceed, that in the wildcat. Rod densities cats is about the same (63 d), birth weight is roughly the same in the area centralis and along part of the horizontal axis in (100 ? 20 gm), and eye opening occurs at the same age (7-10 domestic cats (Steinberg et al., 1973; present results) are higher d after birth). One explanation for the reduction in the size of than those measured in the male wildcat SI. Although it would the domestic cat’s brain is that far fewer neurons are generated be unwise to read too much into this difference, given the vari- early in development. A more counterintuitive, but equally pos- ability in rod and cone densities in other species (Curcio et al., sible, alternative is that the same numbers of neurons are gen- 1990; Wikler et al., 1990; Williams, 199 l), we can certainly say erated in the domestic cat as in the wildcat, but that many more that the rod population is as high (and possibly higher) in the of these cells are subsequently eliminated. For reasons listed domestic cat as in the Spanish wildcat. Similarly, the population below, we think that the decrease in brain size in the domestic of alpha ganglion cells in the domestic cat has not undergone lineage is probably associated with an increase in the incidence any reduction. In fact, densities of alpha cells at the center of of cell death. the area centralis may be slightly higher in domestic cats than Cell death in the domestic cat retina is particularly severe- in the wildcat retinas that we studied. This evolutionary stasis 80% of all ganglion cells die in prenatal and early postnatal in alpha cell number stands in marked contrast to the sharp development (Williams et al., 1986). This magnitude of cell decline in the total ganglion cell population. death exceeds by a wide margin that in other vertebrates in These two intriguing exceptions demonstrate that allometric which ganglion cell death has yet been studied, including rhesus changes in brain mass represent averages that can mask sub- macaque and human (Rakic and Riley, 1983; Provis et al., stantial variation in the magnitude of change among different 1985). As suggested in previous work (Williams et al., 1986) neuron subpopulations (Harvey and Krebs, 1990). As we and the excessive production of ganglion cells in the domestic cat others have shown, these differential effects on neuron popu- fetus may be an ontogenetic trait held over from its large wildcat lations may be prominent even when the evolutionary and ge- ancestor-a form of phylogenetic inertia (Harvey and Purvis. netic separation between two species, or even two strains, is 199 1). The elimination of this surplus may correspond to what exceedingly small (Smith, 1928; Holloway, 1968; Wimer et al., Gliicksmann (195 1) has referred to as phylogenetic cell death. 1976; Ebinger and Lohmer, 1987). In the domestic cat lineage, Our remarkable but admittedly tentative finding that the num- selective pressures appear to have counterbalanced allometric bers of optic axons in wildcat and domestic cat fetuses do not tendencies and thereby have maintained scotopic and motion differ appreciably early in development is consistent with a role sensitivity by leaving intact genetic and epigenetic mechanisms for ganglion cell death operating at a later stage of development responsible for generating and maintaining populations of rods in generating the marked species differences seen at maturity. and alpha ganglion cells. In contrast, populations of cones and The idea that brain evolution may be associated with late beta ganglion cells (particularly in the area centralis) have not fetal and early neonatal modulation in the incidence ofcell death been spared a sharp decline. Cellular selectivity of this sort is also consistent with an observation first made by von Baer serves to emphasize the underlying genetic and epigenetic com- (1828) that early stages of development in related species tend plexity of allometric relations. to be relatively uniform whereas late stages tend to differ (Gould, 226 Williams et al. * Rapid Evolution of the Visual System

1977; Buss, 1987). A developmental and evolutionary shift in Chalupa LM, Lia B (1991) The nasotemporal division of retinal gan- the onset or duration of cell death in a lineage may be just as glion cells with crossed and uncrossed projections in the fetal rhesus monkey. J Neurosci I 1: 19 I-202. effective in changing CNS structure as a heterochronic shift in Chalupa LM, Williams RW, Henderson Z (1984) Binocular interaction the onset or duration of cell proliferation (Goldowitz, 1989; in the fetal cat regulates the size of the ganglion cell population. Wikler and Finlay, 1989). Of course, the modulation of cell Neuroscience 12: I 139-l 146. death is ultimately limited by the degree of neuron overpro- CITES (199 I) Convention on international trade in endangered species duction. Consequently this mechanismcannot provide an ex- of wild flora and fauna, appendices I-III. Endangered species con- vention, Oct. I, 1991, 50 CFR 23.23; Endangered and-threatened planation for the long-term phyletic increase in brain size that wildlife and plants, July 15, 1991, 50 CFR 17.1 1. is prominent in some mammalian lineages. For instance, it is Collier G, O’Brien SJ (1985) A molecular phylogeny of the Felidae: improbable that the brain of Homo sapiensis larger than that immunological distance. Evolution 39:473487. of its Australopithecine ancestorssimply becausefewer neurons Coouer ML. Pettiarew JD (1979) The decussation of the retinothal- amic pathway in the cat, with a note on the major meridians of the are lost at a late stageof development. cat’s eye. J Comp Neurol 187:285-3 12. Our argument implies that the severity of cell death is gen- Clutton-Brock J (I 98 I) Domesticated animals from early times. Cam- erally elevated in the CNS of domestic cats in comparison to bridge: Cambridge UP. that in the CNS of Spanishwildcats-a testable prediction. It Curcio CA, Sloan KR, Kalina RE, Hendrickson AE (1990) Human may be the casethat large litter size and the high metabolic load photoreceptor topography. J Comp Neurol 292:497-523. Darwin C (1890) The variation of animals and slants under domes- of late pregnancy on smaller domestic females serve as a bio- tication, 2d ed,’ rev. New York: Appleton. . chemical brake that eliminates many neurons when they first Davis DD (1962) Allometric relationships in lions vs. domestic cats. become active. This metabolic interpretation provides some Evolution 16:505-5 14. rationale for the loss of normally connected neurons (Williams Ebinger P, Lohmer R (1987) A volumetric comparison of brains be- tween greylag geese (Anser anger L.) and domestic geese. J Himforsch and Herrup, 1988; Chalupa and Lia, 199 1; Oppenheim, 199 1). 28:29 l-299. The modulation of rates of cell death rather than of rates of Edinger T (1948) Evolution of the horse brain. Geol Sot Am Mem cell production could be of selective advantage in making pos- 25:1-177. sible rapid and well-matched evolutionary changesin the size Edinger T (1966) Brains from 40 million years of Camelid history. of interconnected neuron populations (Katz and Lasek, 1978; In: Evolution of the forebrain (Hassler R, Stephan H, eds), pp l53- I6 1. Stuttgart: Thieme. Finlay et al., 1987). Seen in this light, the overproduction of Eisenberg JF (198 1) The mammalian radiation. An analysis of trends neurons in a speciesrepresents a concealedsource of variation in evolution, adaptation, and behavior. Chicago: University of Chi- that can be easily exploited in responseto selective pressureby cage. a simple change in the incidence of cell death (Williams and Eisenberg JF, Wilson DE (1978) Relative brain size and feeding strat- egies in the Chiroptera. Evolution 32:740-75 1. Herrup, 1988). There is a great deal of evidence that the mod- Eldredge N, Gould SJ (1972) Punctuated equilibria: an alternative to ulation of cell death is in large part governed by highly flexible phyletic gradualism. In: Models in paleobiology (Schopf TJM, ed), trophic feedback mechanisms(Purves, 1988). Such responsive pp 82-l 15. San Francisco: Freeman, Cooper. and contingent mechanismsmay effectively loosen the grip of Finlay BL, Wikler KC, Sengelaub DR (1987) Regressive events in epigeneticconstraints on brain evolution (Katz and Lasek, 1978; brain development and scenarios for vertebrate brain evolution. Brain Behav Evol 30: 102-I 17. Hamburger and Oppenheim, 1982). These ideasare compatible Gliicksmann A (195 1) Cell deaths in normal vertebrate ontogeny. Biol with the neatly matched 30% difference in both retinal ganglion Rev 26:59-86. cell and LGN neuron populations that we have documented in Goldowitz D (1989) Cell allocation in mammalian CNS formation: this study. evidence from murine interspecies aggregation chimeras. Neuron 3: 705-713. Modes of development are subject to selection both at the Gould SJ (1977) Ontogeny and phylogeny. Cambridge, MA: Harvard level of the individual and at the level of the species.The pres- UP. ence or absencewithin a lineageof a mechanisminvolving the Gould SJ, Eldredge N (1986) Punctuated equilibrium at the third stage. overproduction and subsequentloss of neurons via trophic in- Syst Zoo1 35:143-148. teractions may provide a concrete developmental basisfor dif- Guillery RW (1966) A study of Golgi preparations from the dorsal lateral geniculate nucleus of the adult cat. J Comp Neurol 128:21- ferences in rates of speciation (Purves, 1988) and, consequently, 50. for selection at, or even above, the species level (Stanley, 1975; Haltenorth T (1953) Die Wildkatzen der Alten Welt. Eine Ubersicht Raff and Kaufman, 1983; Vrba, 1983). iiber die Untergattung Fe/is. Leipzig: Akademische Verlagsgesell- schaft, Geest and Portig. References Hamburger V, Oppenheim RW (1982) Naturally occurring neuronal death in vertebrates. Neurosci Comm l:39-55. Armstrong E (1982) Relative brain size and metabolism in mammals. Harvey PH, Krebs JR (1990) Comparing brains. Science 249:139- Science 220: 1302-l 304. 146. Bekker JM, Stevens CF (1990) Two different ways evolution makes Harvey PH, Purvis A (1991) Comparative methods for explaining neurons larger. Prog Brain Res 83:3745. adaptations. Nature 35 1:6 19-624. Benveniste RE (1985) The contributions of retroviruses to the study of mammalian evolution. In: Molecular evolutionary genetics Hemmer H (1976) Gestation period and postnatal development in (Maclntyre RJ, ed), pp 359417. New York: Plenum. felids. In: The world’s cats. Vol 3. No 2 (Eaton RL. ed). LID 90-100. Seattle: Carnivore Research Institute, Burke Museum, Ut%ersity of Boycott BB, WBssle H (1974) The morphological types of ganglion cells of the domestic cat’s retina. J Physiol (Lond) 240:397--119. Washington. Braendgaard H, Gundersen HJG (1986) The impact of recent stereo- Holloway RL (1968) The evolution of the primate brain: some aspects logical advances on quantitative studies of the nervous system. J of quantitative relations. Brain Res 7: I2 l-l 72. Neurosci Methods 18:39-78. Hughes A (1975) A quantitative analysis of the cat retinal ganglion Buss LW (1987) The evolution of individuality. Princeton: Princeton cell topography. J Comp Neurol 163: 107-l 28. UP. Jacobs DS, Perry VH, Hawken MJ (I 984) The postnatal reduction of Carroll RL (1988) Vertebrate paleontology and evolution. New York: the uncrossed projection from the nasal retina in the cat. J Neurosci Freeman. 4~2425-2433. The Journal of Neuroscience, January 1993, 13(l) 227

Jerison H (1979) Brain, body and encephalization in early primates. Provis JM, van Driel D, Billson FA, Russell P (1985) Human fetal J Hum Evol 8:615-635. optic nerve: overproduction and elimination of retinal axons during Jerison HJ (1973) Evolution of the brain and intelligence. New York: development. J Comp Neurol 238:92-101. Academic. Purves D (1988) Body and brain. A trophic theory of neural connec- Kaas JH (1987) The organization and evolution of neocortex. In: tions. Cambridge, MA: Harvard UP. Higher brain function: recent explorations of the brain’s emergent Purves D, Rubin E, Snider WD, Lichtman J (1986) Relation ofanimal properties (Wise SP, ed), pp 347-378. New York: Wiley. size to convergence, divergence, and neuronal number in peripheral Katz MJ, Lasek RJ (1978) Evolution of the nervous system: role of sympathetic pathways. J Neurosci 6: 158-163. ontogenetic mechanisms in the evolution of matching populations. Radinsky LB (1973) Evolution of the canid brain. Brain Behav Evol Proc Nat1 Acad Sci USA 75: 1349-l 352. I l:l69-202. Kirby MA, Chalupa LM (1986) Retinal crowding alters the mor- Radinsky LB (1975) Evolution of the felid brain. Brain Behav Evol phology of alpha ganglion cells. J Comp Neurol 25 1:532-541. 11:214-254. Koch C, Poggio T, Torre V (1982) Retinal ganglion cells: a functional Radinsky LB (198 1) Brain evolution in extinct South American un- interpretation of dendritic morphology. Philos Trans R Sot Lond gulates. Brain Behav Evol 18: 169-l 87. [Biol] 298:227-264. Raff RA, Kaufman TC (1983) Embryos, genes, and evolution. The Kurten B (I 954) Observations on allometry in mammalian dentition: developmental-genetic basis ofevolutionarychange. New York: Mac- its interpretation and evolutionary significance. Acta Zoo1 Fenn 76: millan. l-122. Rakic P, Riley K (1983) Overproduction and elimination of retinal Kurten B (1958) A differentiation index, and a new measure of evo- axons in the fetal rhesus monkey. Science 2 19: 144 l-l 444. lutionary rates. Evolution 12: 146-157. Reiner A (199 1) A comparison of neurotransmitter-specific and neu- Kurttn B -(1959) Rates of evolution in fossil mammals. Cold Spring ropeptide-specific neuronal cell types present in the dorsal cortex in Harbor Svmo Ouant Biol 24:205-2 15. turtles with those present in the isocortex in mammals: implications Kurten B (1965,) The Camivora of the Palestine caves. Acta Zool for the evolution of isocortex. Brain Behav Evol 38:53-9 1. Fenn 107: l-74. Reinoso-Suirez F (1961) Topographischer Himatlas der Katze fur KurtCn B (1965b) On the evolution of the European wild cat, Felis experimental-physiologische Untersuchungen. Darmstadt: E. Merck. silvestris Schreber. Acta Zoo1 Fenn 1 I I: l-29. Rodriguez de la Fuente F (1979) El gato mantes. In: Fauna ibtrica y Kurten B (I 97 I) The age of mammals. New York: Columbia UP. europea, Vol VI, Bosque caducifolio. Bosque de coniferas y taiga, pp Kurt& B (1988) On evolution and fossil mammals. New York: Co- 60-69. Barcelona: Salvat Editores. lumbia UP. RShrs M (1955) Vergleichende Untersuchungen an Wild- und Haus- Lande R (I 979) Quantitativegenetic analysis ofmultivariate evolution katzen. Zoo1 Anz 155:53-69. aDDlied to brain : bodv size allometrv. Evolution 33:402-4 16. Rosen CD, Harry JD (1990) Brain volume estimation from serial Latimer HB (1938) The weights of spinal cord and of the eyeballs in the adult cat. J Comp Neurol 68: Methods 35: 115-l 24. 395-404. Rowe MH, Dreher B (1982) Functional morphology of beta cells in Lia B, Williams RW, Chalupa LM (1986) Does axonal branching the area centralis of the cat’s retina: a model for the evolution of contribute to the overproduction of optic nerve fibers during early central retinal specializations. Brain Behav Evol 2 1:l-23. development of the cat’s visual system. Dev Brain Res 25:296-301. Rowe MH, Stone J (1976) Properties of ganglion cells in the visual Lindemann W, Rieck W (1953) Beobachtungen bei der Aufzucht von streak of the cat’s retina. J Comp Neurol 167:99-l 26. Wildkatzen. Z Tierpsychol 10:92-l 19. Schindewolf OH (1950) Grundfragen der Palaontologie. Stuttgart: Madarasz M, Gerle J, Hajdu F, Somogyi G, Tomb61 T (1978) Quan- Schwarzerbart. titative histological studies on the lateral geniculate nucleus in the Shatz CJ (1983) The prenatal development of the cat’s retinogeniculate cat. II. Cell numbers and densities in several layers. J Himforsch 19: pathway. J Neurosci 3:482499. 159-164. Smith LA (1928) A comparison of the number of nerve cells in the Mann MD, Glickman SE, Towe AL (1988) Brain/body relations among olfactory bulbs ofdomesticated albino and wild Norway rats. J Comp myomorph rodents. Brain Behav Evol 3 1:111-l 24. Neurol45:483-499. Marschall LG, Corrunccini RS (1978) Variability, evolutionary rates, Stanford LR (1987) X-cells in the cat retina: relationships between and allometry in dwarfing lineages. Paleobiology 4: 101-I 19. the morphology and physiology of a class of cat retinal ganglion cells. Martin PS, Klein RG, eds (1984) Quaternary extinctions: a prehistoric J Neurophysiol 58:940-964. revolution. Tucson: University of Arizona. Stanley SM (I 975) A theory of evolution above the species level. Proc Martin RD, Harvey PH (1985) Brain size allometry. Ontogeny and Nat1 Acad Sci USA 72:646-650. phylogeny. In: Size and scaling in primate biology (Jungers WL, ed), Stanley SM (1979) Macroevolution. Pattern and process. San Fran- pp 147-l 73. New York: Plenum. cisco: Freeman. Menner E (I 939) Vergleichende Untersuchungen iiber die Retina wild- Steinbere RH. Reid M. Lacv PL (1973) The distribution of rods and lebender und domestizierter Caniden. Z Naturwiss Med Grundlagen- cones h the retina of the cat (Fe& dohesticus). J Comp Neurol 148: forsch 93:77-88. 229-248. Mountcastle VB (1980) Medical physiology, 14th ed. St. Louis: Mosby. Stone J (1978) The number and distribution of ganglion cells in the Nikara T. Bishop PO, Pettiarew JD (1968) Analysis of retinal corre- cat’s retina. J Comp Neurol 180:753-772. spondence by studying receptive fields of binocular single units in the Stone J, Rapaport DH, Williams RW, Chalupa LM (1982) Uniformity cat striate cortex. Exp Brain Res 6:353-372. of cell distribution in the ganglion cell layer of prenatal cat retina: Oppenheim RW (I 99 1) Cell death during development of the nervous Implications for mechanisms of retinal development. Dev Brain Res system. Annu Rev Neurosci .l4:453-501. 2123 l-242. Otsuka R, Hassler R (1962) Uber Aulbau und Gliederung der corti- Van Valen L (1969) Evolution of communities and late Pleistocene calen Sehsuhare bei der Katze. Arch Psvchiatr Z Nervenkr Ges Neurol extinctions. In: Proceedings ofthe North American Paleontology Con- Psychiatr 203~2 12-234. vention, Chicago, pp 469-485. Pagel MD, Harvey PH (1989) Taxonomic differences in the scaling von Baer KE (1828) Uber Entwickelungsgeschichte der Thiere. Beo- of brain on bodv weiaht amona mammals. Science 224: 1589-l 593. bachtung und Reflexion, Pt I. Konigsberg: Gebriidem Bomtrager Peichl L, Wassle H (11981) M&hological identification of on- and (available at Beineke Library, Yale University). off-centre brisk transient (Y) cells in the cat retina. Proc R Sot Lond Vrba ES (1983) Macroevolutionary trends: new perspectives on the [Biol] 212:139-156. roles of adaotation and incidental effect. Science 221:387-389. Perry VH, Linden R (1982) Evidence for dendritic competition in the WPssle H, Riemann RJ (1978) The mosaic of nerve cells in the mam- developing retina. Nature 297:683-685. malian retina. Proc R Sot Lond [Biol] 200:441-46 1. Preuss TM (1993) The role of neurosciences in primate evolutionary Wassle H. Levick WR. Cleland BG (1975) The distribution of the biology: historical commentary and prospectus. In: Primates and their alpha type of ganglion cells in the cat’s retina. J Comp Neurol 159: relatives in phylogenetic perspective (MacPhee RDE, ed). New York: 419438. Plenum, in press. Wassle H, Peichl L, Boycott BB (198la) Morphology and topography 228 Williams et al. * Rapid Evolution of the Visual System

of on- and off-alpha cells in the cat retina. Proc R Sot Lond [Biol] the optic nerve of the cat: evidence of size groupings and regional 212:157-175. organization. J Neurosci 3: 1554-l 564. Wgssle H, Boycott BB, Peichl L (1981b) Morphology and mosaic of Williams RW, Herrup K (1988) The control of neuron number. Annu on- and off-beta cells in the cat retina and some functional consid- Rev Neurosci 11:423453. erations. Proc R Sot Lond [Biol] 2 12: 177-l 95. Williams RW, Rakic P (1988a) Elimination ofneurons from the rhesus Wayne RK, Benveniste RE, Janczewski DN, O’Brien SJ (1989) Mo- monkey’s lateral geniculate nucleus during development. J Comp lecular and biochemical evolution of the Camivora. In: Carnivore Neurol 272:424-436. behavior, ecology, and evolution (Gittleman JL, ed), pp 465494. Williams RW, Rakic P (1988b) Three-dimensional counting: an ac- Ithaca, NY: Cor%ell UP. curate and direct method to estimate numbers of cells in sectioned Weber AJ, Kalil RE (1983) The percentage of intemeurons in the material. J Comp Neurol 278:344-352. dorsal lateral geniculate nucleus of the cat and observations on several Williams RW, Bastiani MJ, Chalupa LM (I 983) Loss of axons in the variables that affect the sensitivity of horseradish peroxidase as a cat optic nerve following fetal unilateral enucleation: an electron mi- retrograde marker. J Camp Neurol 220:336-346. croscopic analysis. J Neurosci 3: 135-146. Wikler-KC, Finlay BL (1989) Developmental heterochrony and the Williams RW, Bastiani MJ, Lia B, Chalupa LM (1986) Growth cones, evolution of species differences in retinal specializations. In: Devel- dying axons, and developmental fluctuations in the fiber population ooment of the vertebrate retina (Finlav BL. Sengelaub DR. eds)., DD__ of the cat’s optic nerve. J Comp Neurol 246:32-69. 227-246. New York: Plenum. . . - Williams RW, Borodkin M, Rakic P (I 99 I) Growth cone distribution Wikler KC, Perez G, Finlay BL (1989) Duration of retinogenesis: its patterns in the optic nerve of fetal monkeys: implications for mech- relationship to retinal organization in two Crisitine rodents. J Comp anisms of axon guidance. J Neurosci 1 I : IO8 l-l 094. Neurol 285: 157-l 76. Wimer RE, Wimer CC, Vaughn JE, Barber RP, Balvanz BA, Chemow Wikler KC, Williams RW, Rakic P (1990) Photoreceptor mosaic: CR (I 976) The genetic organization of neuron number in Ammon’s number and distribution of rods and cones in the rhesus monkey horns of house mice. Brain Res I 18:2 19-243. retina. J Comp Neurol 297:499-508. Wong ROL, Hughes A (1987a) The morphology, number, and dis- Williams RW (199 1) The human retina has a cone-enriched rim. Vis tribution of a large population of confirmed displaced amacrine cells Neurosci 6:403-l06. in the adult cat retina. J Comp Neural 255: 159-l 77. Williams RW. ChaluDa LM (1982) Prenatal development ofthe retino- Wane. ROL. Hughes A (I 987b) Develooine. neuronal oonulations of collicular projections in ;he cat: an anterograde transport study. thecat reiinalganglion~ cell laber. J Comp&eurol 262’3473-495. J Neurosci 2:604-622. Zeuner FE (1963) A history of domesticated animals. New York: Williams RW, Chalupa LM (1983) An analysis ofaxon caliber within Harper and Row.